Ideas for building
better software.

Spring and Security In The Times Of AI

Why NSOpenPanel Can't See Your Android Device

All tests run on an 8-year-old MacBook Air. When I was building HiyokoMTP, I hit a limitation that seems obvious in hindsight: you can't use a native macOS folder picker to browse an Android device over MTP. NSOpenPanel — the standard macOS file dialog — has no idea your Android device exists. It only sees what Finder mounts, and Finder doesn't mount MTP devices as a proper volume. So the entire native folder selection API is useless for this use case. macOS exposes the file system to apps throu

Reflections on AI (large language models) in various fields

List of articles on the impact of AI(LLMs) on a developer's life from various perspectives, including: research, students, beginners, professionals, and others: professional: https://jpain.io/god-damn-ai-is-making-me-dumb/ https://ky.fyi/posts/ai-burnout Company: https://mastrojs.github.io/blog/2026-05-23-is-AI-causing-a-repeat-of-frontends-lost-decade/ Research: https://www.moderndescartes.com/essays/ai_and_expertise/ Academy https://htmx.org/essays/yes-and/ https://vickiboykis.com/2026/05/28/

RAG - Hybrid search and RAG pipeline using FAISS DB

Hybrid Search Hybrid search is a combination of dense embeddings and sparse embeddings. Dense embeddings focus on semantic meaning, while sparse embeddings focus on exact keyword matching. By combining both approaches, hybrid search improves retrieval accuracy and relevance. OpenSearch is commonly used as a search engine for: Log analysis Observability and monitoring One of the key features of OpenSearch is hybrid search, which combines: Vector search (dense retrieval) BM25-based search (spars

Why ROU Asset Depreciation Checks Fail When Schedule And Journal Logic Drift

The depreciation problem usually gets reported as a journal problem. Someone sees a depreciation expense that looks off. The export no longer ties cleanly. The journal layer gets blamed first. In practice, the bug is usually upstream by the time anyone opens the journal tab. Most IFRS 16 depreciation failures are drift failures. The schedule still looks plausible. The closing ROU asset is still in the right ballpark. What broke was one of the connections underneath it: the liability was remeasur

Tossed The T.V. but Kept the Remote? Here's how to NOT generate more E-Waste.

This is a submission for the GitHub Finish-Up-A-Thon Challenge Link to the Repo What I Built So what is this? Doesn't look like an API 👁️👁️ Tossed The TV Kept the Remote (TTVKTR) is a simple solution to the common problem of ending up with leftover Infrared(IR) Remotes from appliances long after the appliances are gone. They should be trivially easy to use with whatever you want, instead of ending up in landfills. After finding an old Apple TV Remote in the archi

The Architecture Decisions That Actually Mattered: Building a Production-Ready Multi-Service Backend

What most system design articles skip is the part where you explain why the boring choice was the right one I built a platform that runs social media giveaway events, a gift card marketplace, and telecom gift vending — all in one system, all sharing the same PostgreSQL database and Redis instance, all running in production on infrastructure that costs €27 a month. This article is not about the product. It is about the seven architectural decisions I made while building the backend, why I ma

Frontier Logic at Local Speed: The 2026 Strix Halo Ultimate Benchmark Suite

The era of choosing between "Small & Fast" or "Large & Slow" for local AI is ending. With the release of the Qwen 3.6 family and architectural breakthroughs in inference engines, we can now run frontier-class reasoning on personal hardware at human-reading speeds. In this technical audit, we benchmark the AMD Strix Halo (Radeon 8060S) using a custom-tuned llama.cpp stack to identify the optimal configuration for sovereign intelligence. Our test host ("Stark") utilizes the Strix Halo architecture